From Blanket Denial to Case-by-Case Judgment

How BIS is reframing risk for advanced computing items

This week, I found myself revisiting a part of the regulatory landscape that often sits just outside the day-to-day focus of compliance teams in financial institutions: export controls.

More specifically, the Bureau of Industry and Security (BIS) has revised its license review policy for advanced computing commodities destined for China and Macau. Until now, the default posture for exporting advanced computing items such as high-end GPUs such as H200, MI325X, and comparable products to China and Macau was essentially a presumption of denial. The revised policy moves away from that blanket approach and instead introduces case-by-case licensing review.

This change reflects a more mature and targeted regulatory stance. Rather than assuming every transaction is unacceptable, BIS is signalling that approvals may be possible, but only where exporters can demonstrate a very high degree of confidence around who the end user is, what the equipment will be used for, and how diversion risks are managed.

Export controls start to look a lot like other compliance frameworks

The key message in between the lines is that export control compliance increasingly demands the same level of operational discipline we expect in financial crime compliance: KYC depth, data validation, escalation protocols, and defensible judgment.

The revised policy places heavy emphasis on:

  • enhanced security controls,

  • rigorous customer due diligence,

  • verification of end users and end uses,

  • ongoing monitoring of how the items are deployed.

Strip away the hardware context, and you could easily mistake this for an AML, sanctions, or third-party risk framework.

The end user is now the centre of gravity

Another important shift is how explicitly the policy focuses on end-user scrutiny.

The traditional export control question used to be: Is this item controlled? Is the destination restricted?
The emerging question is: Who exactly will be using this capability, for what purpose, and under what safeguards?

This shifts the burden upstream.

Exporters must now demonstrate not just contractual assurances, but a genuine understanding of:

  • corporate structures,

  • beneficial ownership,

  • data access rights,

  • downstream deployment environments,

  • and potential links to military, surveillance, or restricted AI applications.

I believe this is where many organisations will feel the strain. These are not checks you can complete once at onboarding and file away. They require continuous confidence in your counterparts, not just static representations.

A quiet convergence with AI governance

It’s also worth noting what this change does NOT do.

The revised policy does not replace the broader AI diffusion framework or national security controls around artificial intelligence. Instead, it complements them by tightening the gate around the physical infrastructure that makes advanced AI possible.

Seen through that lens, the move makes sense. If advanced models are increasingly regulated, it follows that the compute enabling those models will attract similar scrutiny. What’s new is how explicitly BIS is tying access to compute hardware to process quality, not just destination or product classification.

Why this matters beyond exporters?

For banks, investors, and firms financing or supporting technology supply chains, this development should not be ignored.

Advanced compute hardware now sits at the intersection of:

  • export controls,

  • national security,

  • AI governance,

  • third-party risk,

  • and sanctions-adjacent compliance.

That means counterparties involved in manufacturing, distribution, financing, leasing, or even cloud provisioning may find themselves subject to deeper questions, not just from regulators, but from their own partners and service providers.

The days when export control compliance could be neatly siloed away from enterprise risk management are clearly fading.

A closing thought

Moving from blanket denial to case-by-case review puts into focus the idea of controls, judgment, and accountability. It rewards firms that can demonstrate real understanding of their customers and end users, and it exposes those relying on box-ticking or contractual boilerplate.

In that sense, the direction of travel feels familiar. Across sanctions, AML, AI governance, and now export controls, regulators are converging on the same expectation:

Show us that you understand the risk and can manage it in practice.

As always, if this topic is landing on your desk as well, I’m happy to compare notes. These developments may look technical, but their implications travel far beyond trade compliance.

Thanks for reading,
Alexey